Repeater Production for the North Atlantic Link By H. A. LAMB* and W. W. HEFFNER* (Manuscript recrived September 20, 1956) Production of submarine telephone cable repeaters, designed to have a minimum trouble-free life of iwenlij years, required many new and refined manufacturing procedures. Care in the selection and training of personnel, manufacturing environment, inspection, and testing, were of great impor- tance in the successful attainment of the ultimate objective. Although quality of product has always been of major significance in Western Electric Company manufacture, building electronic equipment for use at the bottom of the ocean, where maintenance is impossible and replacement of apparatus extremely expensive, required unusual manufacturing methods. MANUFACTUHING OBJECTIVE Late in 1952, the manufacture of flexible repeaters for the North Atlantic Link of the transatlantic submarine telephone cable system was allocated to the Kearny Works of Western Electric Company. In accordance with established practice in initiating radically new products and processes, production of these repeaters was assigned to the Engineer of Maiuifacture Organization rather than to regular manu- facture in the telephone apparatus shops. The job — to produce 122 thirty-six channel carrier repeaters and 19 equalizei-s capalile of operat- ing satisfactorily at pressui'es up to 6,800 pounds per square inch on the ocean floor, with minimum maintenance, for a period of at least twenty years. Initial delivery of repeaters was reciuired for March, 1954, less than a year and a half after the project started. GENERAL PHILOSOPHY Quality has always been the piime consideration in producing appara- tus and equipment for the Bell System. There is an economical breaking point, however, beyond which the return does not warrant the abnormal • Western Electric Company. 103 lO-t THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 expenditures rc(iuircd to approach theoretical perfection. The same philosophy applies to all manufactured commodities, be they auto- mobiles, airplanes or telephone systems. In general, all of these products are physically available for preventive and corrective maintenance at nominal cost. With electronic repeaters at the bottom of the ocean, main- tenance is impo.ssible and replacement would be extremely expensive. The general philosophy adopted at the inception of the project was to build integrity into the product to the limit of practicabihty. To do this, a number of fundamental premises were established, which form the foundation of all operations involved: 1. Manufacturing environment would be provided which, in addition to furnishing a desirable place to work, could be kept scrupulously clean and free from contamination. 2. The best available talent would be screened and selected for the particular work involved. 3. Wage payments would be based on day work, rather than on an incentive plan basis, because production schedules and the complexity of the operations did not permit the high degree of standardization essential to effective wage incentive operation. 4. A sense of individual responsibility would be inculcated in every person on the job. 5. Training programs would be established to thoroughly prepare supervisors, operators, and inspectors for their respective assignments before doing any work on the project. (i. Inspection, on a 100 per cent basis, would be established at every point in the process which could, conceivably, contribute to, or affect the integrity of the product. PREPARATION FOR MANUFACTURE Manufacturing Location It appeared desirable to set up manufacture in a location apart from the general manufacturing area. Experience gained to date has satisfied us that this was the correct approach, since it provided a number of advantages: 1. Administration has been greatly facilitated by having all necessary levels of supervision located in the immediate vicinity of the work. 2. It was necessary for the people on the job to acquire and maintain a new philosophy of perfection in product, rather than a high output at an "acceptable quality level." This was easier at a separate location, since only one philosophy was followed throughout the plant. FLEXIBLE HEPEATEK AL-iNUFACTUEE 105 3. Engineering, production control, service and maintenance organi- zations were located close to actual production and had no assignments other than the project. 4. The small plant, due to its semi -isolation, tends to produce a very closely knit organization and good teamwork. A large number of manufacturing locations were examined and the one selected was a one-story modern structure in Hillside, New Jersey, which provided a gross area of 43,700 square feet. The entire plant was air conditioned ; in most cases, the temperature was controlled to minimum 73 degrees F, maximum 77 degrees F. The air was filtered through two mechanical and one electrostatic filters. Relative humidity was maintained at maximum 40 per cent in all but one area ■ — the capacitor winding room — in which it was necessary to maintain maximum 20 per cent humidity to avoid mechanical difficulty with capacitor paper. While mo.st of the air was recirculated, the air from the cafeteria, cleaning room, locker and toilet rooms was exhausted to the OUTOOOR ENCLOSURE FOR — COMPRESSED GASES A,B,C,D,E INDICATE CLASSIFICATION OF AREA Fig. 1 ^ Plant layout. 106 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 outside atmosphere. Two separate air conditioning systems were in use. One, of 300 tons capacity, provided for most of the plant, while a smaller unit of 30 tons capacity served the capacitor winding, testing, and im- pregnating rooms. Each installation had its own air filtering and condi- tioning equipment. Plant Layout The plant layout is illustrated in Fig. 1. All working areas, with the exception of the repeater enclosure area, were individually enclosed, and walls from approximately four feet above the floor were almost entirely of reinforced glass. This arrangment facilitated supervision by other than first-line supervisors, who were located with the groups, and provided a means of viewing the operations by the many visitors at Hillside, without contaminating the critical areas or disturbing the operators. Analysis of Design for Facilities and Operations In analyzing the design for manufacture there were, of course, numer- ous instances where conventional methods and facilities were entirely adequate for the job. Since their inclusion would contribute little to this article, we shall confine the description to those cases which are new or unusual. Collaboration with Bell Telephone Laboratories in Preparation of Manu- f adoring Information Early in 1953 a coordination committee was established, consisting of representatives from the various Laboratories design groups and Western engineers, which met on a bi-weekly basis during the entire period preceding initial manufacturing operations. These meetings provided a clearing house for questions and policies of a general nature for this particular project and served to keep all concerned informed as to the progress of design and the preparations for manufacture. It is customary, during the latter stages of development of any project at the Laboratories, for Western engineers to participate in the prepara- tion of manufacturing information as an aid in pointing the design to- ward the most economical and satisfactory production methods and facilities. Since the decision to use the Bell System repeater in the Trans- atlantic system was based on the performance of the Key West-Havana installation, and the fact that changes in design would require further FLEXIBLE REPEATER MAXUFACTURE 107 trials over an extended period of time, only minor changes to facilitate manufacture were made. Further, since some experience had been gained by the Laboratories in producing repeaters for that installation, it was decided to "pool" effort in preparing the manufacturing process informa- tion, which is normally Western's responsibility. Close cooperation of the two groups, therefore, has resulted in the production of repeaters wliich are essentially replicas of those in the initial installation except for the internal changes necessary to increase transmission capacity from 24 to 36 channels. Other Western Electric Locations and Outside Suppliers During the development work on the Key West-Havana repeaters, the Hawthorne Works of Western Electric had furnished the molyb- denum-permalloy cores for certain inductors, the Tonawanda Plant had furnished mantlrelated resistance wire, and the AUentown Plant had fabricated the glass seal subassemblies. Since the experience gained in this development work was extremely "\'aluable in producing the additional material required for the Transatlantic system and since the facihties for doing the work were largely available, these various locations were asked to furnish similar material for the project. Although the Kearny Crystal Shop had not been involved in the Key West-Havana project, arrangements were made there to make the crystals for this project, since facilities were available, along with considerable experience in producing precision units. Subcontracted Operations While it was believed, initially, that all component parts for repeaters should be manufactured by Western Electric, critical analysis indicated that it was neither desirable nor economical in certain cases. One of the outstanding examples in this category is the hardened and ground chrome-molybdenum steel rings that constitute the strength members in the repeater and sustain the pressures developed on the ocean bottom. Purchasing the many large and varied machine tools and associated lieat treating eciuipment necessary to produce these parts would have re(iuired a substantial capital expenditure and additional manufacturing space. Arrangements, therefore, were made with a highly qualified and well equipped supplier to produce the rings, using material furnished by Western, which liad been pre\-iously inspected and tested to very strin- gent requirements. lOS THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 The situation attending the manufacture of a relatively small number of comparatively large copper parts used in the rubber and core tube seals was much the same. Here, again, the large size machine tools and additional manufacturing space, required for only a short time, would have increased the o^'er-all cost of the project considerably. These parts, therefore, were subcontracted in the local area and inspection was per- formed by Hillside inspectors. A safeguard, in so far as integrity is concerned, was provided by the fact that these were individual parts that could be reinspected at the time of delivery. No subassembly operations that might possibly result in oversight of a defect, were subcontracted. Manufacturing Conditions Two major problems confronted us in planning the manufacture of repeaters. First, to produce units that were essentially perfect; and second, to prevent the contamination of the product by any substance that might degrade its performance over a long period of time. In ap- proaching both of these objectives, it was realized that the product had a definite economic value which the cost of production should not exceed. In many cases, therefore, it was necessary to rely on judgment, backed by considerable manufacturing experience, in determining when the "point of no return" had been reached in refining processes and practices. The initial approach to this phase of the job was to classify, with the collaboration of Bell Telephone Laboratories, all of the manufacturing operations involved as to the degree of cleanliness required. In setting up these criteria, it was necessary to evahiate the importance of contami- nation in each area and the practicability of eliminating it at the source or to insure that whatever foreign material accumulated on the product was removed. A representative case is the machining of piece parts. While the shop area is cleaner, perhaps, than any similar area in industry, the very nature of the work is such that immediate contamination cannot be avoided since material is being removed in the form of chips and turn- ings, and a water soluble oil is used as a coolant. In this instance, however, the parts can be thoroughly cleaned end their condition observed liefore leaving the area. Conversely, in the case of an operation such as the assembly of paper capacitors into a container which is then hermetically sealed, it is vitally necessary to insure that both the manufacturing FLEXIBLE KEPEATER MANUFACTURE 109 area and the processes are free from, P.nd not conducive to producing, particles of material which are capable of causing trouble. The various classifications established for the production areas include specific requirements as to temperature, relative humidity, static pres- sure with respect to adjacent areas, cleanliness in terms of restrictions on smoking and the use of cosmetics and food, and the type and use of special clothing. Special Clothing Employees' clothing was considered one of the most important sources of contamination for two reasons; first, for the foreign material that could be collected upon it and carried into the manufacturing areas, and second, that various types of textiles in popular use are subject to con- siderable raveling and fraying. After considerable study of many types of clothing for use in critical areas, the material adopted was closely woven Orion, which has proved to be acceptably hnt-free. The complete uniform — supplied at no cost to employees — consists of slacks and shirts for both male and female employees, Orion surgeon's caps for the men and nylon-visored caps for the women, hi addition shoes, without toecap seams, were provided. Nylon smof^ks were furiii.shed to protect the uniforms while employees moved from locker rooms to the entrance vestibule. Two changes of clothing were provided each week, and the laundering was done by an outside concern. Employees to whom this special clothing was issued were paired for locker use. Both kept their uniforms and special shoes in one locker and their own clothes and shoes in the other. This prevented the transfer to the uniforms of any foreign material that might exist on the street clothing. At the entrance vestibule to the A, B, and C areas (Fig. 1) the employees were required to clean their shoes in the specially designed facilities provided and to wash their hands in the wash basins installed for this purpose. Smocks were then removed and hung on numbered hooks that line the walls at the end of the vestibule. Employees were then permitted to go to their work positions within the inner areas. At any time that it was necessary for employees to leave the work areas for any purpose, they were required to put on their smocks in the vesti- bule and upon their return, to go through the cleaning procedure again. Employees in the other areas were provided only w^th smocks, mainly for the protection of their clothes since the work involved could soil or stain them but could not be contaminated from the clothing. 110 THE BELL SYSTEM TECHNICAL JOUENAL, JANTJAEY 1957 Cleaning Schedules were established for cleaning the areas at regular intervals, the frequency and methods depending upon the type of manufacturing operations and the activity. Usually, the vinyl plastic floors were ma- chine scrubbed and vacuum dried. Walls, windows and ceihngs were cleaned by hand with hnt-free cloths. Manufacturing facihties such as bench tops, which were linoleum covered, were washed daily. Test sets, cabinets, test chambers and bench fixtures were also cleaned daily. Hand tools were cleaned at least once a week by scrubbing with a solu- tion of green soap, rinsing in distilled water, followed by alcohol and then dried in an oven. Dust Count Since it was impossible to determine what contaminating material in the form of air-borne particles might be encountered from day to day, and what the effect might be during the life of the repeaters, the general approach to this problem was to control, so far as possible, the amount of dust within the plant. In order to verify, continuously, the over-all effectiveness of the vari- ous preventive measures, dust counts were made in each classified area at daily intervals, using a Bausch and Lomb Dust Counter. This device combines, in one instrument, air-sampling means and a particle-counting microscope. Over a two-year period it has been possible to maintain, in certain areas, a maximum dust count of between 2,000 and 3,500 particles per cubic foot of air with a maximum size of 10 microns. Control checks, taken outside the building at the employees' entrance, generally run upwards of 25,000 particles per cubic foot, a good portion of which are of comparatively large size. PRODUCTION" AND PERSONNEL Equipping the plant, obtaining and installing facilities, and selecting and training personnel proceeded on a closely overlapped basis with receipt and analysis of Bell Telephone Laboratories' product design information. Because of the critical nature of the product, provisions were made not only for the most reliable commercially available utilities and services, but also for emergency lighting service in some areas. Maintenance and service staffs had to be built up rapidly as the super- visory and manufacturing forces were being developed. FLEXIBLE REPEATER M.\JS'UFACTUHE 111 "Qualification" of All Personnel Before employecM were assigned to production work they were re- quired to pass a qualification test established by the inspection organi- zation to denionsli'ate satisfactory performance. Programs were, there- fore, set up for "vestibule" training and qualification of new employees. This activity was carried on by full-time instructors who had been trained by Western and Bell Laboratories engineers. Training was carried out in two stages: 1. (a) The employee received instruction and became acquainted with equipmeut and lequirements. (b) A practice period in which the employee developed techniques aiid worked under actual operating (conditions, with all work submitted to regular inspection. 2. A qualification period in which the employee was recjuired to demonstrate that work satisfactory for project use could be produced. The main objective during stage 1 was progressive quality improve- ment and in stage 2 the maintenance of a satisfactory quality'' level over an extended period of time. Employees made a definite number of units at acceptable quality levels in order to qualify. The number of units required for training varied with the type of work and the ease with which it was mastered. All personnel were required to pass qualification tests before being assigned to production work anfl were restricted to that work unless trained and qualified for other work. Employees trained on more than one job were requalified before being returned to a previous assignment. Records of the performance of individual operators started in the training stage wore continued after the employees were assigned to production work. The performance record of the operators was based on results obtained during the inspection of their work, while that of the inspectors was liased on special quahty accuracy checks of their work. Personnel Selection It was apparent that the new manufacturing techniques, including the cleanliness and quality demands, would necessitate that all shop supervisors and employees be very carefully selected. It also appeared (and this was subsequently confirmed) that after the careful selection and training of supervisors, long training periods would be required for specialty selected shop employees. In selecting first line shop supervisors, such factors as adaptability, personality, and ability to work closely with the engineers were of para- mount importance. For the parts and apparatus included in their re- 112 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 sponsibility, they were required to thoroughly learn the design, the operations to be performed, the faciUties to be used, the data to be recorded, the cleanliness practices to be observed — and in most cases, prepare themselves to be able to do practically all of the operations, be- cause subsequently they had to train selected operators to perform criti- cal operations to very high quality standards under rigidly controlled manufacturing conditions. As shop supervisors and employees were assigned to the manufacture of repeaters, they were thoroughly indoc- trinated in the design intent and the new philosophy of manufacture. Standard ability and adaptabihty tests were used in a large number of cases to assist in proper selection and placement of technicians. Tests for finger and hand dexterity; sustained attention; eyes, including per- ception and observation ; and reaction time of the right foot after a visual stimulus. (The latter test was relatively important for induction brazing operations.) Other requisite considerations were a high degree of de- pendability and integrity, involving intellectual honesty and conscien- tious convictions; capability of performing tedious, frustrating, and exasperating operations against ultra-high quality standards, verifying their own work; perseverance and capability to easily adapt to changes in assignment and occupation or the introduction of design changes. We considered whether or not they would stand up under "fishbowl" operations, wherein they would receive a considerable amount of ob- servation from high levels of Western Electric Company and Kell Sys- tem management and other visitors. Also, could they duplicate high quality frequently after qualifying for a particular operation? During the period of repeater manufacture, the number of employees rose from less than 50 in January, 1954, to a maximum of 304 by Feb- ruary, 1955, after which there was a gradual reduction to a level of about 205 employees for six months and then a gradual falling off as we were completing the last of the project. In the period from May to December, 1954, between 30 and 45 employees were constantly in training prior to being placed on productive work. During 1955 this decreased to prac- tically no employees in training during the midpart of the year and there- after training was required merely to compensate for a small labor turn- over and employee reassigim.ient. It is significant that labor turnover was very low and attendance was exceptionally good during the life of the Hillside operations. Personnel Training The original plan, which was generally followed, was to prove in the tools for each phase of the job, followed by an intensive program of train- FLEXIBLE REPEATER MANUFACTURE 113 ing. Iiuloctriiuitiou of laboratory technicians could be considered as "vestibule triiining" in that they were acclimated to the area and con- ditidUH, given oral instruction in the work, then given practice materials and demonstrations and, when cjualified, were started on making project material. To do this, extra supervisors were required at the beginning of the job. A super\'isor trained a few employees, qualified some of them, and began work on the project. Another supor\isor was then required to train additional employees who, as they became qualified, were trans- ferred to the supervisor re.sponsible for making project apparatus. Addi- tional testing of the employees, instruction and reinstruction and, in some cases, retraining were required. In practically all cases, we were able to fit an employee selected for work at Hillside into some particular group of operations. The extra emphasis on selection and training cre- ated a well-balanced team that later resulted in considerable flexibility. During all of this training our supervisors worked closely with engineers and inspectors who understood the design intent and the degree of per- fection required. At the bogimiing, each technician was trained for only one operation of a particular job, such as (1) winding Type X capacitors or (2) im- pregnating all paper capacitors or (3) winding Type Y transformers and so became an expert on this one operation. Later, the tours of duty for many technicians were broadened to cover several operations. Comimmications To keep employees informed, wc occasionally assembled the entire group, presenting informati\'e talks on current production plans and our future l)usiness prospects. Motion pictures were shown of the cable laying ships and the operations of cable sphcing and cable laying. A dis- play l)oard, showing all of the repeater components, was mounted on the wall of the cafeteria. This informed the operators just where the parts were used in apparatus; also, just where their products went into the wired repeater unit, and how all electrical apparatus was enclosed against sea pressure in the final repeater. In small groups, all of the em- ployees at Hillside were given a short guided tour of the plant to see the facilities and hear a de.scription of the operations being performed in each area. These communications were extended to everyone at the Hillside Plant, iticluding those who did not work directly on the product. It was our conviction that the maintenance men, boiler operators, oilers, station wagon chauffeur, janitors, and clerical workers in the office were all interested and could do a better job if kept informed of the needs and progress of the project. 114 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 Scheduling Capacity was provided at the Hillside Shop to manufacture a max- imum of 14 repeaters in a calendar month. This envisioned 6-day opera- tion with some second and third shift operations; due allowance was made for holidays and vacations, so that the annual rate would be ap- proximately 160 enclosures per year. (An enclosure is either a repeater or an equalizer.) Some of the facilities and raw materials were ordered late in 1953. This ordering expanded early in 1954 and continued through 1955 to include parts to be made by outside suppliers and the parts and appara- tus to be made at Hillside. Apparatus designs were not all available at the beginning of the job, and the ultimate quantities required were also subject to sharp change as the project shaped up, thus further compli- cating the scheduling problem. Because of the time and economic factors involved, coupled with the developmental nature of the product and processes, one of the most difficult and continuing problems was the balancing of production to meet schedules. For this task, we devised "tree charts" for the apparatus codes and time intervals in each type of repeater or equalizer for each project. Each chart was established from estimates of the time required to accomplish the specified operations and the percentage of good prod- uct each major group of operations was expected to produce. RAW MATERIALS Many of the specifications were written around the specific needs of the job and embodied requirements that were considerably more strin- gent than those imposed on similar materials for commercial use. As a result, it was necessary for many supphers to refine their processes, and, in some eases, to produce the material on a laboratory basis. One example is the container, or repeater enclosure, which consists, in part, of a seamless copper tube approximately If inches in diameter having a ^-inch wall and approximately 8 feet long. This material was purchased in standard lengths of 10 feet. The basic material was re- quired to be phosphorous deoxidized copper of 99.80 per cent purity. The tubing, as delivered, had to be smooth, bright, and free from dirt, grease, oxides (or other inclusions including copper chips), scale, voids, laps, and shvers. Dents, pits, scratches, and other mechanical defects could not be greater than 0.003 inch in depth. The tubing had to be concentric within 0.002 inch and the curvature in a 10-foot length not exceed ^ inch to facilitate assembly over the steel rings. FLEXIBLE REPEATER MANUFACTURE 115 Only one supplier was willing to accept orders for the tubes, and only on the basis of meeting the mechanical requirements on the outside sur- face. To establish a source of supply, it was necessary to accept the sup- plier's proposal on the basis that some of the tubes produced could be expected to meet requirements on the inside as well as the outside sur- face. Inspection of the inside surface was performed with a 10-foot Bore- scope. The suppher then set aside, overhauled, and cleaned a complete group of drawing facilities for this project. In addition, a number of refinements were made in lul)rication and systematic maintenance of tools. After all refinements were made and precautions taken, however, the yield of good tubes in the first 400 produced was less than 1.0 per cent. Consulta- tions with Western and Bell Laboratories' engineers, and with the sup- plier's cooperation, raised the yield to approximately 50 per cent. Procurement of satisfactory mica laminations for capacitors intro- duced an unusual problem. The best grade of mica available in the world market was purchased which the supplier, under special plant condi- tions, spht and processed into laminations. Despite care in selection and processing, only 50 per cent of the 250,000 laminations purchased met the extremely rigid requirements for microscopic inclusions and delam- inations, and less than 8 per cent survived the capacitor manufacturing processes. A large number of the parts, and the most complex, are made from methyl-methacrylate (Plexiglass). At the time manufacture began, there was little, if any, experience or information a^'ailable on machining this material to the required close tolerances and surface finish. Consequently, considerable pioneering effort was expended in this field before satis- factory results were obtained. The methacrylate parts cover a wide range of size and complexity — • from l^-inch diameter by 4|-inch long tubular housing to tiny spools |-inch diameter and ■^-inch long. Most of the parts are cyhndrical in shape with some semi cylindrical sections that must mate with other sections to form complete cylinders. Others have thin fins, walls, flanges and projections. Five representative parts are shown in Fig. 2. Methyl-methacrylate has a tendency to chip if tools are not kept sharp and care is not used in entry or exit of the tool in the work, particularly in milling. In some cases, it is necessary, with end-milling, to work the cutter around the periphery of the area for a slight depth so that sub- sequent cuts will not break out at an unsupported area. Normally, with a sharp cutter and a 0.010-inch finish cut, and a slow feed, chipping will not result. High-speed steel tools with zero rake were used for turning 116 THE BELL SYSTEM TECHNICAL JOrHNAL, JANUARY 1957 Fig. 2 — Methyl methacrylate parts. and boring operations. Standard high-speed milling cutters and end mills were used for milling except for the cutting edges, which are honed to a fine finish. A clearance angle of 7 degrees for milling and 10 degrees to 15 degrees for lathe work was found moat satisfactory. In lathe work, the general rule was light feeds (0.003 inch-0.005 inch) and small depth of cut. However, the depth of cut could be safely varied over a wide range depending upon many factors, such as type of part, quality of finish, machine and tool rigidity, effective application of coolant, and tooling to support and clamp the part. In one operation of boring a 1^-inch diameter by 4|-inch deep blind hole within ±0.002 inch, the boring terminates in simultaneously facing the l)ottom of the hole square with its axis. A cut ^-inch deep with a light feed was taken with a spe- cially designed boring tool with the coolant being fed through the shank to the cutting edge. All completely machined parts were annealed for 12 hours at 175° F. HIGHLIGHTS IN ASSEMBLY AND BKAZING Repeater units are encased in hardened steel rings which previously had been tested at 10,000 pounds per square inch hydraulic pressure. This pressure is approximately 50 per cent higher than the greatest pres- sure expected at ocean bottom. The steel rings were encased in a copper sheath and closed at each end with a glass-to-Kovar seal, with the cen- tral conductor coming through the glass to the outside. The copper sheath was then shrunk to the steel rings and glass seals using 6000 pomids per FLEXIBLE KEPEATEH iLiXUFACTURE 117 square inch hydraulic pressure, aucl the glass seal was then high-fre- quency brazed to the copper sheath. To keep the ocean bottom pressure off the glass seals and also to ter- minate the cable insulation, a rubber seal is brazed in to the copper con- tainer tube adjacent to each glass seal. This rubber seal consists of rub- ber bonded to l)rass, which has been brazed to the copper portion of the seal. The rubber terminates in polyethylene through five steps of com- pounds containing successively less rubber and more polyethylene. The polyethylene can be readily bonded by molding to the polyethylene in- sulation of the cable. The central conductor passes through a central brass tube in the rubber seal, which is also bonded to the loibber. To protect the rubber seals from the deleterious effects of salt water immersion for long periods of lime, a copper core tube is brazed over each rubber seal. The core tube is arranged to equalize the pressure in- side and out when submerged at ocean bottom pressure. This is accom- plished with a bulge of ncoprene filled with polyisobutylene, on the far end of the core tube, which transmits the pressure to the inside of the core tube seal. To make doubly sure that no salt water reaches the rubber seal, a copper cover is brazed into the container outside the core tube connector on each end. This cover is also brazed to the core tube connector. The interatice between each of the above four seals is filled with polyiso- butylene, which is viscous and inert and has very good insulating qualities. Each end of the repeater closure (Fig. 3) contains five successive brazed joints. Any one of these ten brazes, if not perfect, could cause the loss of the repeater closure and jeopardize the entire repeater. All of these brazes were made with the repeater in u vertical position to insui'e an even di.stribution of the brazing alloy fillet around the jomt. An upending device was provided at the pit brazing location to raise the repeater on its carrier to a ^-ertical position with either end up and move it into position for brazing. The repeaters were brought into the brazing area on an overhead monorail and an electric hoist. The shorter repeater assemblies, before core tube and cable stub assembly, were up- ended by hand and brazed from a raised platform. It was necessary to make all of these brazes bj' high-frequency induc- tion heating, since the heat must be intense, contained within a ^-ery narrow band, evenly distributed, and the area protected from oxidation by a somewhat reducing atmosphere. The heat must be very intense since the time interval for the shortest braze was 10 seconds maximum and the longest was 30 seconds. A large part of the heat was dissipated lis FLEXIBLE REPEATER HL\NUFACTUHE 119 by being conducted at a high rate from the copper parts to the water in the cooling jackets used to contain the heat in a very narrow band. Circulating cooling water within a jacket prevented heat from being conducted down the copper container tube to the preceding seals or to the repeater unit. This water-cooled jacket was positioned only f inch below the inductor, and the water was in intimate contact with the con- tainer tube, which Is sealed off at both ends with rubber "0" rings. In addition, for the glass seal braze, the glass inside the seal cavity was kept covered with water during the heat cycle. The water was fed in and si- phoned out to a constant level which was kept under observation by the operator and the inspector to make sure that the glass was covered at all times. The rubber seal was also water jacketed on the inside of the seal to pre^'ent deleterious effects of the heat on the rubber insulation around the central conductor. The imier cover braze was quenched before the 10-second maximum interval had expired to insure that the heat did not penetrate to the polyisobutylene at a sufficient rate to deteriorate it or the rubber inside. Distribution of the heat around the container tube at the braze area was controlled by locating the work in the inductor so that the color came up essentially evenly all the way around and at the proper level to bring a fillet up to the top of the braze joint within the allowable time limit. The time limit was determined by experiment so that none of the previously assembled parts were damaged by the heat. This determina- tion of the proper heat pattern and the prevention of overheating re- quired the development of considerable skill on the part of the operator. The variables encountered made it essential to rely on an operator to control the heat rather than to utilize the timer with which the induction heating equipment is normally controlled. The area to be heated for brazing was protected from oxidation by enclosing it in a separable tran.sparont plastic box and flooding the in- terior with a gas consisting of 15 per cent hydrogen and 85 per cent nitro- gen. This atmosphere is somewhat reducing and not explosive. The brazing surfaces of the parts were chemically cleaned immediately before assembly and extreme care was exercised to keep them clean until brazed. The container tube was shrunk to the respective glass, rubber, core tube, and cover seals using hydraulic pressure so that the surfaces to be brazed and the brazing alloy were in intimate contact within the brazing area. If the parts were clean and kept from oxidizing by the protective atmosphere, the alloy would flow upward by capillary action and form a fillet around the top of the seal, impervious to any leak. The braze in each case was then leak tested with a heUum mass-spec- 120 THE BELL SYSTEM TECHNICAL JOUHNAL, JANUARY 1957 trometer type leak detector. A gas pressure of helium at least 25 per cent greater than the maxknum pressure to be encountered at ocean bottom was used. In addition, a radioisotope was used to test the effectiveness of the final tabulation pinch welds and o\'ei-brazes which were kept open for the leak tests under high pressure helium. These tests were made with water pressure about 25 per cent greater than the maximum ocean bottom pressure. The completed repeater was inserted in a chamber 80 feet long; the chamber was then filled with water and the pressure raised to 7,500 pounds per square inch and held at that pressure for at least 15 hours. At the end of this period the closure had to show no sign of crushing or leaking. The repeater unit sealed in the closure must be extremely dry to func- tion properly. Any water vapor which might remain after the closure is sealed, or enter during the estimated 20-year minimum life, must be scavenged. A sealed desiccator with a thin diaphragm was, therefore, assembled into the repeater unit sections. After completely drying and seahng the repeater unit except for one tubulation, the diaphragm of the desiccator was ruptured by dry nitrogen pressure and with the en- closure filled wdth dry nitrogen the final tubulation was immediately sealed off. To insure that the diaphragm was actually broken, a micro- phone was strapped to the outside of the repeater over the location of the desiccator and a second microphone arranged at the end of the closure to pick up background noises. A pen recorder was used to record the sound from the two microphones and also the change in nitrogen pressure. Three simultaneous pips on the chart gave definite indication that the diaphragm had ruptured and that the desiccant had been ex- posed to the internal atmosphere of the repeater. QUARTZ CRYSTAL UNITS MANUFACTURED AT KEARNY The primary purpose of the crystal unit is to provide the means of identifying and measuring the gain of each repeater in the cable. This basic crystal design is in common usage. The exacting specifications for this apphcation, however, imposed many problems and deviations from normal crystal manufacturing processes. Raw Quartz was specially selected for this crystal unit. The manu- facturing process of reducing the quartz to the final plate followed the recognized methods through the roughing operations. Due to the rigid end requirements, the finishing operations were performed xmder labora- tory conditions. Angular tolerances were one-third of normal limits. No evidence of surface scratches, chipped edges or other surface imper- FLEXIBLE REPEATER MANUFACTURE 121 fectioiis visible under 30X magnification were permitted. This resulted ill a process shrinkage five times that experienced in normal crystal plate manufacture. In this use, the crystal units were required to meet performance tests at currents as low as one-thousandth of a microampere — far below the current \'alues usually encountered. Improved soldering techniques had to be developed for soldering the gold plated phosphor bronze and nickel wires used, because it was found that the electrical performance of the units was directly related to the ([uality of soldered connections. Although one-seventh of Western's production of quartz crystal units iU'e in glass enclosures, the applicable techniques in glass working re- (juired a complete revision. Glass components such as the stem and bulb purchased from established sources were found to be far below the stand- ard required for this crystal unit. For example, the supplier of the glass tubing used in the manufacture of .stems was required to meet raw mate- rial specifications that embodied coefficient of thermal expansion, soften- ing point of glass, density, refractive index, and volume resistivity. The glass stems made from this tubing by regular manufacturers were found unacceptable and the processes used by these sources could not be readily adapted to meet the desired specifications. The glass stems contained four lead wires made from 30-mil Grade "A" nickel wire butt welded to 10-mil light borated Dumet wire. To assure the quality of the metal to glass seal, each wire was inspected under SOX magnification for tool marks and other surface imperfections. The finished stem assemblies were inspected under 30X magnification for dimensions, workmanship, cleanliness and minute glass imperfections, then individually stored in a sealed plastic envelope. The glass bulb in this crystal unit is known as the T921 design com- monly used in the electron tube industry. The high quaUty required, howe\'er, made TOO per cent inspection necessary. Examination under ;iOX magnification resulted in rejected bulbs for presence of scratches, open bubbles, chips and stones. Physical limits for inside and outside diameters as well as wall thickness were causes for additional rejects. Only one per cent of the commercial bulbs were found acceptable, and these were also stored in a sealed plastic envelope. The final major assembly operation consisted of seahng the glass bulb to the stem which had had the crystal sub-assembly welded to the nickel wires. The techniques for "sealing in" used in quartz crystal or electron tube manufacture were unsuited. Two important factors in this crystal unit, which required the de\'elopment of new processes, were the prox- imity of soft soldered connections to the sealing fires and the demands 122 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 O ■ pH FLEXIBLE REPEATER MANUFACTURE 123 that the glass seal contain a minimum of residual tensile stress. These two problems were resolved collectively by performing the sealing opera- tion on a single spindle glass seaUng machine. Accm^ate positioning of the glassware and sealing fires, together with precise timing and tem- perature controls, achieved the desired results. Evaluation of residual stresses were made by inspections using a polarimeter and by a thermal shock test. The maximum safe stress was established at 1.74 KG/mm^. The thermal shock test required successive immersion of the unit in Ijoiling water and ice water. The electrical char- acteristics of these units exceeded all others made previously by Western Electric. The ratio of reactance to effective resistance ("Q") was greater than 175,000 — twice that ever previously produced and 17 times that required in the average filter crystal. Stability for frequency and resistance was assured by a 28-day aging test. During this period, precise daily resonant frequency and resistance measurements were recorded against temperature within 0.1° C. The maximum permissible change was 0.0005 per cent in frequency and +5 per cent to — 10 per cent in resistance. GLASS SEALS MANUFACTURED AT ALLENTOWN The glass seal used to close each end of the container for the repeaters and equalizers is manufactured at the AUentowii Works of the Western Electric Clompany. The unit is essentially a glass bead-type seal. It insulates the central conductor of the repeater from the container and serves as a final vapor barrier between the cable and the interior of the repeater. As such, it backs up several other rubber and plastic barriers as shown in Fig. 3. Fig. 4 shows the various components, subassembhes, and a cross-sec- tion of the unit. The unit consists of the basic seal brazed in the Kovar outer shell, to which is brazed, a copper extension provided with two brazing-ring grooves. One of these grooves is used in brazing the seal, along with support members, into a length of container tubing in the same manner as the seal is ultimately brazed into the repeater. Packaging of the seal in this manner was necessary to pressure test the seal. Under test, in a specially constructed chamber 10,000 psi of helium gas pressure was applied to the external areas of the packaged glass seal and a mass spectrometer type leak detector was connected through the tubulation to the internal cavity of the packaged unit. In this manner, the interface of the glass to metal seal, the brazed joints, and the porosity of the metal were checked for leakage. The unit is left in this package for delivery to provide protection during shipment. Before the seal could be used, 124 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 it was machined from the package by cutting the copper extension to length, leaving the second groove for use in brazing the seal to the re- peater and removing the container tubing and the support members. The basic seal consists of the cup, central conductor and glass. The cup (smaller cylindrical item in the upper lefthand corner of Kig. 4) was machined from Kovar rod. The wall of the cup is tapered from a thickne-ss of 0.025 inch at the base to 0.002 inch at the lip. The last 0.000 inch of the lip is further tapered from this 0.002 inc-h to a razor edge. The internal surface is better than a 63-micro-inch turned finish and was also liquid honed to give it a uniform matte finish. The central con- ductor (slim piece in the upper right-hand corner of Fig. 4) was also machined from Kovar rod. Both the cup and central conductor were further processed by pickling, hypersonically Cleaning in deionized water, and decarburizing. The glass, a borosilicate type of optical quality, was cut from heavy walled tubing. The glass tubing was hand polished, lapped and etched to remove surface scratches, and to arrive at the spe- cified weight. It was also fire polished and hypersonically cleaned to remove all traces of surface imperfections and to assure maximum clean- liness. In order to make the basic glass seal, the metal parts had to be oxidized under precisely controlled conditions. For the oxidizing operation, a suitable fixture was loaded mth brazed shell-cup assembhes, central conductor assemblies, and a Kovar disc, which had been prepared in precisely the same manner as the cups and central conductors. The disc was carefully weighed before and after oxidizing and the increase in weight divided by the area involved yields the weight gain due to oxida- tion for each run. Limits of 1.5 to 2.5 milligrams per square inch of oxide were set. This operation was performed by placing the loaded, sealed retort, through which passed a metered flow of dried air, into a furnace for a specified time-temperature cycle. In the glassing operation the oxidized shell assembly, the carbon mold and the central conductor were placed in a fixture and held in the proper relation.ship. The carbon mold served to support the glass, while it was being melted, in that section between the cup and central conductor where the glass was normally unsupported. The prepared cut glass tubing was loaded into the Kovar cup and the fixture was sealed into the retort. During the glassing cycle, a constant flow of nitrogen passed through the retort to provide an atmosphere which minimized any reduction or further oxidation of the already carefully oxidized parts. After the proper purging period, the retort was placed in the furnace. In the furnace, the glass melted and formed a bond with the oxidized Kovar of the cup and FLEXIBLE REPEATER MANUFACTURE 125 centi'al conductor to form the seal. After the specified temperature-time cycle, the retort was removed from the furnace, allowed to partially cool and then placed into an annealing oven. Vertical furnaces and retorts were used for brazing, decarburizing, oxidizing and glassing. By varying the type of gases flowing into the retorts, atmospheres which are reducing, oxidizing, or neutral were ob- tained. To provide maximum uniformity of process, separate retorts and holding fixtures were provided for operations involving hydrogen and for air-nitrogen operations, so that a retort or a fixture used for hydrogen treatments was never used for oxidizing or glassing. PILOT AN]) REGULAR PRODUCTION We called our first efforts Practice Parts and Training; the next we called Pilot Production. Next, certain items identified as Trial Laying Repeaters and Oscillators were manufactured for use in "proving in" the ship laying gear. To prove in manufacturing facilities, a few un- equipped housings were made without the usual electrical components normally in a repeater. Similarly, each of the apparatus components and parts required exploratory and pilot effort before regular production could be undertaken. As might be expected, the manufacturing yield of components meeting all requirements was very low during the early stages of the undertaking. However, substantial improvement was brought about as experience was gained. Comments on some of the production problems, highlights, and yield results, follow. Paper Capacitors were manufactured only after painstaking qualifying trials and tests had been performed on each individual roll of paper. Cycling and life testing, procurement of acceptable ceramic parts and gold-plated tape and cans, selection and matching of rolls of paper for winding characteristics, and similar problems, all had to be completely resolved to a point of refinement previously unattempted for telephone apparatus. Composite percentage yield for all operations on paper capacitois is shown in Fig. 5. Yield is shown as the ratio of finished units of acceptable quality to the number of units started in manufacture. Afira Capacitors were made from only the most meticulouslj' selected laminations, as mentioned earlier. Even the best mica is particularly susceptible to damage in processing. In spite of experience and knowl- edge of this, the multiple handling of the laminations contributed an unusually high material shrinkage as each separate lamination needed to be cleaned, then handled individually many times through the proc- 126 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 75 70 \i\/ i , i^ly 65 -MlvL \ Q ili 55 >- AMll n/ Til V Ul O Ul 0. 40 t^L /I / /' 25 20 / 1 II 1^ 1 1 1 1 1 1 1 1 1 1 1 1 JFMftMJJASONDJFUAMJJASONDJFMAMJJASONO 1954 1955 1956 Fig. 5 ^ Paper capacitor yield. esses. The art of silk screening was applied to deposit silver paste in a specific area or areas on each side of a lamination. A sharply defined rectangular area was required so that when superimposed one over an- other the desired capacitance would be obtained. Cementing of mica laminations onto machined methacrylate forms presented some addi- tional problems through the bowing of the mica laminations as the ce- ment cured. Obtaining screens that would give the proper length and width dimensions for the coated area, was another problem. A silk screen woven of strands of silk obviously hniits, by the diameter of the threads, the extent to which the dimensions of an opening may be increased or decreased. Beryllium copper U-shaped terminals were used to clamp the layers of mica together into a stack. Control of the pressure used in crimping these terminals was found to be very critical in view of the exceptionally tight limits on capacitance and stability. Fig. 6 shows the composite yield at various times for all mica capacitors. Resistors, There were three designs of ceramic resistors, which were resistance -wire wound on ceramic spools. These were intended to be assembled into the hole inside the core tube on which the paper capaci- tors were wound. Special winding machines equipped with binocular FLEXIBLE REPEATER MANUFACTURE 127 80 75 70 65 60 u a. 40 35 30 25 20 K H NO PRODUCTION II I I I I I [ I I I I I I I I I I I I I [ I I I JFMAUJJAS0NDJFMAMJJAS0NDJFMAMJJA30ND 1954 1955 1956 Fig. 6 — Mica capacitor yield. attachments were necessary to wind these resistors. Other resistors were hand wound on methyl-methacrj'late forms, or on the outside of the ceramic containers, for certain types of paper capacitors. Rough adjust- ments were required of the lengths of resistance wire prior to winding, and close adjustments to resistance values were mude after the ^\'indings were completed and before leads were attached to resistors. Again it was necessary to provide periodic samples that could be placed on life test by the Laboratories to ascertain that the manufacturing processes were under control. The.se .samples, in all po.ssible cases, were taken from prod- uct that would normally be rejected becau.se of some minor defect, but which would not in any way detract from the validity of the life tests. The making of hard solder splices between nichrome resistance wire and gold-plated copper leads, and keeping ceramic parts from coming in contact with metal surfaces and thereby being contaminated because of the ceramic's abrasive characteristics, were two major problems on resistors. Fig. 7 indicates I'esistor yields. Inductors comprised 20 different designs, most of which were air core, but there were some for which it was necessary to cement permalloy dust cores into pockets of the methacrylate form, and thereafter using 128 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 JFMAMJJAS0N0JFMAUJJA30NDJFMAMJJAS0NO 1954 1955 1956 Fig. 7 — Resistor yield. wire on a shuttle, wind by hand the turns rciiuired to produce an in- ductor. The.se varied from a very small inductor, smaller in diameter than a pencil, to a fairly large "figure eight" inductor with turns having a major diameter of about Ij inches. Each layer of a winding was in- spected with a microscope to insure that the wire had not been twisted or kinked, or that the insulation was damaged or uneven. Some of the shuttles became fairly long so that they could hold the amount of wire required to make a continuous winding. The operator's handling of this shuttle, as she moved it down around the openings in the methacrylate part, or placed it on a bench to proceed with the interlea^'ing tape, de- manded considerable dexterity and concentration to insure that the shuttle was not turned over — which in effect would put a twist in the wire. Although best known means were used to sort cores for their mag- netic properties prior to the time a winding was made, the limits on the inductors themselves were so close that subsequently a large number of windings were lost. The best cores that could be selected, plus the best winding practice, could not produce 100 per cent of the inductors within the required limits. Crazing of the insulation on the wire; cementing together of two methacrylate parts or of permalloy cores into pockets of FLEXIBLE REPEATER ULANUFACTUHE 129 90 as f\j\ ^^y\cURRENl 1 I r \l Xy ^k--^- 75 \ \J>'-^ V \l "-CUMULATIVE \ '' V o ilj 65 >- \r HI U tr 55 HI 0. 50 45 40 35 30 1 1 1 1 1 1 f 1 1 1 1 1 1 1 1 11 1 1 1 1 1 JFMAUJJAS0NDJFMAMJJA5ONDJFUAMJJASOND 1954 1955 1956 Fig. 8 — Iiidiiclor yield. inethaciylate parts, and handling those inductors having long dehcate leads, were the most troublesome items on this apparatus. Fig. 8 shows manufacturing yield for inductors. Networks combined several codes of component apparatus, such as a mica and a paper capacitor, resistor and an inductor. Six networks were used in each repeater unit consisting of two interstage networks, an input, an output, and two beta networks. They demanded a most deUcate wiring job in that stranded gold-plated copper wires had to be joined in a small pocket in methyl methacrylate, where a minimum amount of heat can be applied; otherwise the methacrylate is affected. After soldering, a minimum amount of mo"\'ement of the stranded wire was permitted, inasmuch as the soldered gold-plated copper wire be- comes quite brittle. Repeater Units, are wired assemblies consisting of seventeen sections in which there are six networks, three electron tubes, one gas tube, one crystal, three high voltage capacitors, one dessicator and two terminal sections. The successive build-up of these materials left little chance to make a repair because a splice in a lead was not permissible. It is during this assembly stage that a repeater received its individual identity be- 130 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 cause of the frequency of the particular crystal assembled into the unit. A manufacturing yield of 100 per cent was achieved in the assembly and wiring of repeater units. It was necessary to calibrate the test equipment for this job very closely. Bell Telephone Laboratories and Western Electric worked at length to calibrate the testing details and the test sets for individual net- works. Adjustments in components apparatus to bring the network to the fine tolerances required were accomphshed by minute scraping of the silvered mica on a mica capacitor or removing turns from wire- womid inductors. The cementing of methacrylatc parts, which was a troublesome item on mica capacitors and inductors, also had to be con- tended with on networks. PACKING AND SHIPPING COORDINATION Repeaters were packed in Western Electric specially designed 34-foot long aluminum containers, weighing 1,000 pounds. Forty of these con- tainers were made by an outside firm. Fig. 9 shows two containers tied down in a truck trailer. The repeaters were nested in a pocket of poly- ethylene bags containing shaped rubberized hair sections in order to cushion the repeaters during their subsequent handhng and transporta- tion. The instrumentation required with each case was tested, properly set, and inspected prior to its use on each outgoing case. The instruments were a shock recorder to register shocks in three planes, and a thermome- ter to register the minimum and maximum temperatures to which the repeater had been exposed. Arrangements were made with a commercial trucking company to provide three specially equipped truck trailers, which could be cooled by dry ice during hot weather and warmed by burning bottled gas during cold weather so as to control temperature within the 20-degree F. to 120-degree F. called for in the repeater spe- cification. Appointment of a shipping coordinator supervisor added tremendously to the smooth functioning of services and provided the continuing vigi- lance required to protect repeaters and deliver them to the right place at the right time. His responsibihty was to coordinate all the shipping information and arrangements from the time the item was ready for packing at the Hillside plant, through all trucking arrangements to the armoring factory, to the airport, to England, and to follow, with sta- tistical data and reports, each enclosure until we were able to record the date on which the repeater was laid or stored in a depot. FLEXIBLE REPEATER MANUFACTURE 131 Fig. 9 — Shipping contaiuera. INSPECTION PLAN AND PHOCEDUKES Gaieral It is axiomatic that quality is not obtained by inspection but must be built into the product. However, the Inspection Organization docs have the responsibihty of certifying that the desired quaUty exists. Our eval- uation indicated that the ordinary inspection "screening" would be inadequate to insure the high degree of integrity demanded and that additional safeguards would have to be provided. These controls were achieved, in a practical way, by: (1) Selective placement, intensive training and subsequent qualifica- tion testing of all persoimel. (2) Inspection during manufacturing operations in addition to in- 132 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 spection of product after completion, and regulating inspection so that critical characteristics received repetitive examination during the process of manufacture and assembly. (3) A maintenance program for inspection and testing facihties which provided checks at considerably shorter intervals than is considered normal. (4) Inspection and operating records and reports that point out areas for corrective measures. (5) Records of quality accuracy for all inspection personnel as an aid in maintaining the high quality level. (6) Verification of all data covering process and final inspection as a certification of the accuracy of these data and that the apparatus satis- factoiily meets all requirements. Selection and Training of Inspection Personnel ^. The quality of a product naturally depends upon the skills, attitude, and integrity of the personnel making and inspecting it. It was realized that in order to develop the high degree of efficiency in the inspection organization necessary to insure the integrity of the product, personnel of very high caliber would be required. These employees would have to be (1) experienced in similar or comparable work, (2) they would have to be precise, accurate and, above all, dependable, (3) in order to reduce the possibihty of contamination and damage they would have to be neat and careful, and (4) they would require the ability to work in harmony with other employees, often as a member of a "team," in an environ- ment where their work would be under constant scrutiny. Most of the inspection employees selected to work at Hillside were transferred from the Kearny Plant and had an average Western Electric sei-vice of twelve years. They were hand-picked for the attributes out- lined above, and the "screening" was performed by supervision through personal interviews supplemented by occupational tests given by the personnel department. These tests, which are in general use, are designed to evaluate background and physical characteristics, and they were given regardless of whether the employee had or had not previously taken them. The following group of tests is an example of those given inspectors and testers of apparatus components : (1) Electrical — ac-dc theory and application. (2) Ortho-Rater — Eye test for phoria, acuity, depth, and color. (3) Finger Dexterity — Ability and ease of handUng small parts. FLEXIBLE REPEATER MANUFACTURE 133 (4) Special — Legibility of handwriting, ability to transcribe data and to use algebraic formulae in data computations. Inspection Plan The general plan of visual and mechanical inspection consisted of: (1) Inspection of every operation performed — and in many cases partial operations — during the course of manufacture. This is of par- ticular importance where the quality characteristics are hidden or inac- cessible after completion of the operation. (2) Repeated inspection at subsequent points for omissions, damage and contamination. (3) Rejection of product at any point where there was failure to ob- tain inspection or where the results of such inspection had not been recorded. Most of the visual inspection was performed at the operators' posi- tions to reduce, to a minimum, the amount of handling that could result in damage and contamination. Visual inspection covered three general categories: (1) Inspection of work after some or all operations had been com- pleted, such as the machining of parts. (2) Inspection at those points where successive operations would cover up the work already perform^ed. An example of this is the hand winding of toroidal inductors where each layer of wire was examined under a mi- croscope for such defects as twists, cracks, and crazes in enamel insula- tion, spacing and overlapping of turns, and contamination before the op- erator was allowed to proceed with another layer. While being inspected, the work remained in the holding fixture, which was hinged in such a man- ner as to permit inspection of both top and bottom of the coil. Inductors received an average of 13 and a maximum of 2G visual inspections during winding. (3) Continuous "over-the-shoulder" inspection, where strict adher- ence to a process was required or where it was impossible to determine, by subsequent inspection, whether or not specific operations had been performed. In these cases, the inspector checked the setup and facilities, observed to see that the manufacturing layouts were being followed, that the operations were being performed satisfactorily, and that spe- cifications were being met. ELECTRICAL TESTING The electrical testing, in itself, was not unusual for carrier apparatus and runs the gamut from dc resistance through capacitance, inductance, 134 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 and effective resistance, to transmission characteristics in the frequency band 20-174 kc. What was unusual were the extremely narrow hmits imposed and the number and variety of tests involved as compared to those usually specified for commercial counterparts. The following two examples will serve to illustrate the extreme meas- ures taken to prove the integrity of the product : (A) One type of Resistor was wound with No. 46 mandrelated nichrome wire to a value of 100,000 ohms plus or minus 0.3 per cent. This resistor received six checks for dc resistance, five for instantaneous stability of resistance and two for distributed capacitance, at various steps in the process which included six days' temperature cycling for mechanical stabilization. This resistor was considered satisfactory, after final anal- ysis of the test results, if: (a) The difference in any two of the six resist- ance readings did not exceed 0.25 per cent, (b) The change in resistance during cycling was not greater than 0.02 per cent, (c) The "instantaneous stability" (maximum change during 30 seconds) did not vary more than 0.01 per cent. In addition, it was required that the distributed capaci- tance, minimum 7, maximum 10 mmf, should not differ from any other resistor by more than 2 mmf. (B) For high voltage paper capacitors, the 0.004-inch thick Kraft paper, which constitutes the dielectric, was selected from the most promising mill lots which the manufacturers had to offer. This selection was based on the results obtained from tests that involve examination for porosity, conducting material and conductivity of water extractions. These tests were followed by the winding and impregnation in Halowax of test capacitors. The test capacitors were then subjected to a direct voltage endurance test at 266 degrees F for 24 hours. Samples of prospective lots of paper, which have passed the above test, were then used to wind another group of test capacitors that were subsequently impregnated with Aroclor and sealed. 1,500-volt dc was then apphed to the capacitors at 203° F for 500 hours. In case of failure, a second sampling was permitted. After the foregoing tests had been passed, the supplier providing the particular mill lot was authorized to sht the paper. Upon receipt, six special capacitors were wound, using a group of six rolls of the paper being qualified. These capacitors were then impregnated, checked for dielectric strength at 3,000-volt dc, and measured for capacitance and insulation resistance. The capacitors were then given an accelerated life test at 2,000-volt dc, temperature 150° F, for 25 days. Each lot of six satisfactory test capacitors qualified six rolls of paper for use. Product capacitors were then wound from approved paper, and the dry units checked for dielectric strength at 300-volt dc. Capacitance FLEXIBLE REPEATER MAISfUFACTURE 135 was checked and units were then assembled uito cans and ceramic covers soldered in place. Assemblies were pressmized with air, through a hole provided for the purpose, while the assembly was immersed in hot water to determine if leaks were present. Capacitors were then baked, vacuuni dried, impregnated, pressurized with nitrogen, and sealed oJT. The com- pletely sealed units were then placed in a vacuum chamber at a tem- perature of 150° F, 2 mm. mercury, for 3 hours to check for oil leaks. Capacitance was I'echecked and insulation resistance measured. After seven days, capacitors were unsealed to replenish the nitrogen that had been absorbed by the oil, resealed and again vacuum leak tested. An X-ray examination was then made of each individual unit to verify internal mechanical conditions. Capacitors were then placed in a tem- perature chamber and given the following treatment for one cycle: 16 hours at 150°F; 8 hours at 75°F; 16 hours at O^F; 8 hours at 75°F. At the end of ten days, or 5 cycles, the insulation resistance and con- ductance was measured and a norm established for capacitance. Capacitors were then recycled for ten days, and, if the capacitance had not changed more than 0.1 per cent, they were satisfactory to place on production life test. If the foregoing conditions had not been met, the capacitors were recycled for periods of ten days until stabilized. At that time, 10 per cent of the capacitors in every production lot were placed on "Sampling Life Test", which consisted of applying 4,000- volt dc in a temperature of 150°F for 25 days. At the same time, the balance of the capacitors in the lot were placed on production life test at 8,000-volt dc in a temperature of 42°F for 26 weeks. At the end of this time, the insulation resistance w^as measured and the capacitance checked at To^F and at 39°F. The difference in capacitance at the two temperatures could not exceed +0.001, —0.005 mf, and the total ca- pacitance could not exceed maximum 0.3726, minimum 0.3674 mf. The capacitance from start to finish of the life test could not have changed more than plus or minus 0.1 per cent. If all of the preceding requirements had been satisfied, the particular lot of capacitors described was considered satisfactory for use. The foregoing examples are typical of the procedures evolved for insuring, to the greatest degree possible, the long, trouble-free life of all apparatus used in the repeater. Radioisolope Test There were many new and involved tests which were developed and applied to the manufacture of repeaters. One of the most unique is the use of a radioisotope for the detection of leaks imder hydrauHc pressure. 136 THE BELL SYSTEM TECHNICAL JOURNAL, JANXJAnY 1957 The initial closure operations consisted of brazing into each end of the repeater housing a Kovar-to-glass seal. These seals are equipped with small diameter nickel tubulations which were used to flush and pressurize the repeaters with nitrogen. After these operations had been performed, one of the tubulations was pinchw^elded, overbrazed and coiled down into the seal cavity. The repeater was then placed in a pres- sure cylinder with the open tubulation extending through and sealed to the test cylinder. A mass spectrometer was then attached to the tubula- tion and the te.st cylinder pressurized with helium at 10,000 psi. At the conclusion of this test the repeater was removed from the test cylinder and, after breaking the desiccator diaphragm, the remaining open tubu- lation was pinehwelded and overbrazed. At this point, it became neces- sary to determine whether the final phichweld and overbrazing would leak under pressure. Since there was no longer any means of access to the inside of the repeater, aU testing had to be done from the outside. This was accom- plisheil by filhng the glass seal with a solution of radioisotope cesium 134, which was retained by a fixture. The repeater was then placed in a test cylinder and hydrauhc pressure applied, which was transmitted to the radioisotope in the fixture. After 60 hours under pressure, the re- peater was removed from the cylinder and the seal drained and washed. An examination was then made with a Geiger counter to determine if any of the isotope had entered the final weld. The washing procedure, after application of the isotope solution, in- volved some sixty operations with precise timing. In the case of the re- peater at the rubber seal stage where both ends were tested, it was desirable that these operations be performed concurrently. This was accomphshed by recording the entire process on magnetic tape which, when played back, furnished detailed instructions and exact timing. RAW MATERLAL INSPECTION As might be expected, raw materials used in the project were very carefully examined and nothing left to chance. Every individual bar, rod, sheet, tube, bottle or can of materials was given a serial number and a sample taken from each and similarly identified. Each sample was then given a complete chemical and physical analysis before each corre- sponding piece of material was certified and released for processing. In many cases, the cost of inspection far exceeded the cost of the material. However, the discrepancies revealed and the assurance provided, more than justify the expense. Detailed records of all raw material inspection were compiled and furnished to the responsible raw material engineer who examined them. FLEXIBLE REPEATER fcL\NUFACTUHE 137 critically, as an additional precaution before the material was released to the shop. INSPECTION RECORDS To eUminate, as much as possible, the human element in providing assurance that all prescribed operations had been performed satisfac- torily, inspected properly and the resuhs recorded, means were estab- lished to compile a complete history of the product concurrent with manufacture. This was accomplished through the provision of permanent data books of semilooseleaf design, which require a special machine for removing or inserting pages. Each of these books covered a portion of the work involved in pro- ducing a piece of apparatus and contained a sequential list of pertinent operations and requirements prescribed in the manufacturing process specifications. Space was provided, adjacent to the recorded information, for both the operator and inspector to affix their initials and the data. A reference page in the front of each book identified the initials with the employees' names. All apparatus was serially numbered and the data were identified accordingly. If a unit was rejected, that serial num- ber was not reused. These data books, in addition to estabUshing a complete record of nianufacture, provided a definite psychological advantage in that people were naturally more attentive to their work when required to sign for responsibility. QUALITY ACCURACY As pointed out previously, every precaution was exercised in selecting and training inspection personnel assigned to the project. However, it was realized at the outset that human beings are not infallible and that insurance, to the greatest degree possilile, would ha\'e to be provided against the probability of errors in observation and jugment. Quahty accuracy evaluation procedures were, therefore, established for deter- mining the accui'acy of each inspector's performance. Quality accuracy checking was performed by a staff of five Inspection Representatives and involved an examination of the work performed by inspectors to determine how accurately it was inspected. Materials which the inspector accepted and those which had been rejected were both examined. VERIFICATION AND SUMMARY OF DATA As an added measure of assurance as to the integrity of the product, procedures were established for verifying and simimarizing the inspec- tion records for each serially numbered component, up to an including complete repeaters. 138 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 Verification involved a complete audit of the inspection records to provide assurance that all process operations were recorded as having been performed satisfactorily, that the prescribed inspections had been made, and that the recorded results indicated that the product met all of the specified requirements. This work was performed by a group of six Inspection Eepresentatives who had considerably experience in all phases of inspection and inspection records. As the verification of a particular piece of apparatus proceeded, a verification report was prepared which, when completed, contained the most pertinent inspection data, such as: (1) Recorded measurements of electrical parameters. (2) Values calculated from measurements to determine conformance. (3) Confirmation that all process and inspection operations had been verified. (4) Identification (code numbers and serial or lot numbers) of mate- rials and components entering into the product at each stage of manu- facture. The verification report usually hsted the data for twenty serial num- bers of a particular code of apparatus along with the specified require- ments. Included, also, was a cross-reference to all the inspection data books involved so that the original data could be located easily. These verification reports were prepared for all apparatus up to and including the finally assembled and tested repeaters. The following gives an indication of the number of items examined in the verification of one complete repeater: Items verified in data books 17 , 593 Items verified on recorder charts 1 , 142 Calculations verified 1 , 580 20,316 Number of entries on verification reports 4 , 070 Verification reports, in addition to presenting the pertinent recorded data, provided a "field" of twenty sets of measurements from which it was easily possible to spot a questionable variation. For example, it was the adopted practice on this project to examine, critically, any charac- teristic of a piece of apparatus, in a universe of twenty, which varied considerably from the rest, despite the fact that it was still within limits. While the number of cases turned up in the verification process which have resulted in rejection of product are relatively few, we believe that the added insurance provided, and the psychological value obtained, considerably outweigh the cost.